Note : Les descriptions sont présentées dans la langue officielle dans laquelle elles ont été soumises.
1~24~67
The present invention relates to the encoding of analog
signals to be transmitted in digital form and more particularly
it relates to an analog-to-digital conversion method and to the
associated converter, wherein the values of successive samples
of an analog signal are converted into a sequence of binary
digits related to those values.
As is known, in order digitally to encode an analog
signal, a sampling is first effected at predetermined time intervals,
thus obtaining a train of amplitude - modulated pulses (PAM); then
10 each pulse is encoded by replacing it with a binary number which
expresses both its amplitude and sign (PCM). The encoding ~
operation is generally effected by recognizing the sign of/PAM
sample and then by comparing the sample with a predetermined number
of current or voltage thresholds having polarity in accordance with
the sign, so as to specify by successive approximations the sample
amplitude value. The num~er of thresholds depends on the degree
of accuracy required for the encoding of the amplitude level of
the PAM sample; greater accuracy reduces the quantization noise,
but req~res a greater number of bits. For instance according to
20 the recommendations issued by the ~nternational Committees on PCM
telephone systems, 8 bits are specified for each sample, the first
representing the sign whi~ the remaining seven designate 128
different amplitude levels.
In known converters, the PAM sample usually first under-
goes sign recognition and then is compared with suitable reference
voltages or currents, which will be hereinafter referred to as
"positive- or "negative thresholds' depending on the result of
sign recognition, thus obtaining either a bit 1 or a bit 0
indicating a positive or negative sign and n bits which are each
30 1 or 0 depending on whether the sample is greater or smaller than
the thresholds with which it has been successively compared; after
coding is complete or while it is in progress, a compression opera-
tion occurs and the bits are reduced from n to seven.
-- 1 --
` 1~24~*7
However for very small amplitude signals, this method is
often a source of errors owing to possible misrecognition of the
sign of the sample; the consequences of such misinterpretation
are comparison with thresholds of the wrong sign and a sequence
of logical zeros at the output of the converter.
The reason for such misrecognition is that sign recognition
is effected as soon as the PAM sample appears in the converter,
before the transients due to sampling expire. Particularly if
the sample has a rather small amplitude, the transients can
cause a rapid sequence of sign inversions giving rise to the
above cited consequences. Moveover, the use of such converters
does not permiL sign recognition in correspondence with the last
portion of the sample to be compared, as its previous recognition
is necessary to choose the sign of the thresholds against which
comparison is to be made.
These disadvantages can be overcome by the analog-to-
digital converter of the present invention, which effects ampli-
tude on coding of the PAM sample before recognizing its sign, thus
allowing greaterimmunity from the errors due to the incomplete
decay of transients, and includes means to transfer the sign bit
to the beginning of the digital word that forms the information
unit of the standarized PCM code.
It is a main object of the present invention to provide
a method for digitally encoding amplitude-modulated pulses in
which the binary words obtained are representative of the sign
and amplitude of the pulse, said amplitude-modulated pulses being
applied simultaneously and independently of their sign to two
analog/digital converters, one for positive pulses and the other
for negative pulses, each providing at its output a bit sequence
of logical zeros or a bit sequence representating an encoding of
the pulse amplitude according to whether the pulse is of ~he
sign for which the converter is designed; said sequences then
being summed bit by bit so as to make available as a single
sequence the amplitude encodings of any pulse regardless of sign
-~ ~12~867
the pulse sign being determined after amplitude encoding of its
amplitude by recognition of the converter that has supplied the
amplitude encoding sequence, said sign in turn being encoded com-
patibly with the amplitude encoding. Preferably the bits of the
binary word comprising the amplitude and sign encoding of the pulse
are stored in parallel and read in series to obtain a desired
arrangement of the bits within the word.
It is a further object of the present invention to provide
means for carrying out the above method.
These and other characteristics of the present invention will
become clearer from the following description of a preferred
embodiment thereof, given by way of example and not in any limiting
sense, taken in connection with the annexed drawings in which:
- Fig 1 is a block diagram of an analog-to-digital converter
in accordance with the invention;
- Fig. 2 is a block diagram of a feedback analog-to-digital
converter using the encoding method of the invention.
In Fig. 1 references two converters CA and CB are designed
respectively for encoding positive and negative PAM sample.~ v
These converters, in general use in digital encoding techniques,
supply an encoded binary number having a certain number of bits
(generally seven) for each PAM sample having a polarity in accor-
dance with that for which the sampler was designed. A sample of
opposite polarity produces an output signal which is a sequence of
similar bits (generally zeros).
The binary words from converters CA and CB are applied by
lines 8 and 9 to a logic adder A, which sums them bit by bit to
provide on line 11 a flow of binary words relating to both
positive and negative PAM samples.
Information as to the sign of the samples is subsequently
added to the bits of the amplitude encoding this sign being de-
termined by recognizing which converter CA or CB has supplied
the information. The logic LS for this task supplies a sign bit
on line 10 to a circuit SS which also receives on line 11 the
amplitude encoding of the PAM sample. This circuit has a set
1124867
of parallel-in, series-out memory registers, which can be read
to provide a predetermined bit sequence on line 6. In this manner,
the binary word present on line 6 can be organized so as to begin
with the sign bit followed by the bits defining the amplitude,
beginning with the most significant.
Fig. 2, shows in greater detail a block diagram of an
analog-to-digital converter of the feedback type, as a particular
example of application of the invention.
Similar references in both Fig. 1 and Fig. 2 denote gen-
10 erally similar blocks and lines upon which the same types ofsignals appear.
Two sets of reference generators CC and CC' provide
current thresholds of positive or negative polarity respectively,
according to the sign of their supply, under control of a device
SC supervising the encoding.
Two terminals 2 and 3 connected respectively to positive
and negative voltage sources provide two opposite current
directions in the two sets of reference generators.
Terminals El, E2,..., Ei,...,En and El', E2',...Ei',
20 ..., En' receive commands by which individual current thresholds
are selected. These commands come from terminals Ti and Ti' of
the supervising device SC.
The blocks CC and CC' contain _ current generators
which can be selected, and in the case of this particular example
are realised by means of resistors, denoted by Rl ... To Rn and
Rl' to Rn', placed in series with diodes, acting as switches and
denoted by references ~1 to Dn, and Dl' to Dn'. Resistors R and R'
are the ground-connected branches of the current generators in
which are summed all the currents, passed by the respective sets
30 CC and CC' of current generators.
The input terminal 1 receives PAM samples from a sampler
of any known type, which are applied through decoupling resistors
RA and R~' and lines 4 and 5 to the blocks CC and CC' respectively,
and also to blocks CS and CS'. These latter blocks are ~hreshold
comparators, which compare the current injected into (or sub-
11;i~4~67
tracted from) their input by the PAM sample and the current sub-
tracted from (or injected into) the same input by those current
generators of sets CC and CC' which are activated at any moment;
a binary signal at the outputs of the comparators denotes the result
of the comparison.
A conventional logic inverter i:fi placed at the output
of threshold comparator CS complements the bits obtained by the
comparison effected in CS; the purpose of this step is discussed
hereinafter, during description of the operation of the converter.
10 The binary signals from IN and CS', present on lines 8 and 9
respectively, are added, bit by bit, by a conventional logic adder
A and are then applied to a logic circuit B, supplying at its output
texminal 6 a flow of data arranged according to the standardized
PCM system, by suitably arranging the relative position of single
bits within logic words.
The supervising device SC receives, on lines 8 and 9,
information as to the result of the comparison effected by the
threshold compara;tors CS, CS' and, depending on this, sends suitable
commands to blocks CC and CC', through the connection (not shown)
20 between terminals Ti and Ti' and terminals Ei and Ei'. Timing
signals for the commands are supplied by a conventional time base,
denoted by BT. Typically, the current threshold value supplied
by CC or CC' is incremented or decremented by an amount that is
halved at each successive comparison, depending on the result
of the comparison itself. The device SC also supplies on line 10
a sign relating to the PAM sample and obtained by recognizing which
threshold comparator has provided encoding of the sample.
The operation of the whole converter of Fig. 2 will
now be reviewed, assuming a PAM sample to be present at the input
30 terminal 1. Through RA and RA', it reaches the two comparators
CS and CS' where it is compared with the current thresholds supplied
by the sets of generators CC and CC'~ Where the PAM sample is
positive, the comparison effected in CS with the positive thre-
sholds always results in a logic 1 at the output of CS, whereas
1~2486'7
in CS', owing to the negative thresholds,the result may be either
logic 1 or logic 0 depending on whether the current injected by
PAM sample is higher or lower than that subtracted by the current
generators selected in CC'.
The comparison of the PAM sample with the thresholds
is effected by means of a feedback method controlled by the super-
vising device SC which sends commands to the terminals Ei, Ei'. The
sample is compared first with the highest threshold, then, accord-
ing to the result of this first comparison, another comparison
10 is effected either with a threshold obtained by adding to the
previous threshold a second threshold having half the amplitude
or directly with said second threshold. The comparisons continue
with the addition or subtraction of successive thresholds each
with an amplitude half that of the previous one, until encoding
of the sample is complete. All these comparisons iake place at
predetermined time intervals, determined by time base BT. The
term feedback type converter derives from this feature that the
re~ult of a previous comparison determines the direction of change
of the threshold current for the following comparison.
In spite of the comparison being effected at the same
time in CS and CS', there is no possibility of error since only
one of the two lines 8 and 9 carries a significant binary word,
whereas the other line oarries a word consisting of a sequence
of logic zeros. By adding in A the result of the encoding effected
by one threshold comparator with the sequence of logic zeros sup-
plied by the other threshold comparator, the desired result is
always obtained whatever the sign of the PAM sample may be.
The comparator CS, when acting upon a positive PAM
sample gives a sequence of logic ones as the voltage at its input
30 terminal, taking into account the sign of the sample as well as
that of the activated threshold, is always positive with respect
to the ground terminal. Likewise, for negative PAM samples, CS'
gives as a result a sequence of logic zeros. Consequently, in
order to render the outputs of the two comparatorscompatible,
112~67
the inverter IN is placed at the output of CS so that whenever
the PAM sample has a sign opposité to the one for which the
comparator CS or CS' has been provided, the binary word present
on the line (8 or 9) to adder A consists only of a sequence of
logic zeros.
The adder A effects bit by bit addition of the signals
present on wires 8 and 9, and provides at its output a sequence
of seven bit words representing in binary code the amplitude of
the PAM samples. The sign of each sample is determined by SC
10 at the end of the encoding, based on recognition of which threshold
comparator furnished the encoding of thesample; if the comparator
was CS (that is the comparator associated with the set CC of
generators fed by a positive supply~ the sample was negative,
and viceversa if the comparator was CS'. For samples so small as
to be unable to supply any bit 1 at the output of the adder A,
sign recognition is effected in the time interval corresponding
to the last bit, without the output of any current generator being
applied to the comparators.
In this case the recognition takes place at a time
20 by which the transients have already decayed and thus the prob-
ability of error is very low. In addition encoding takes place
in an ordered manner beginning from the highest threshold and pro-
gressing through lower and lower thresholds, down to zero for
determination of the sign. The sign information is coded by one
bit which is applied to B on line 10, where it occupies the eighth
and last location in the PCM word. It is the task of logic circuit
B to convert this PCM system, with the sign bit at the end of the
word, into a standardized PCM system, with the sign bit at the
beginning. The circuit B can be realized by means of a parallel-
30 in, serial-out regis~er, the serial readout beginning with the sign
bit.